Track LLM Spend in n8n Workflows + Budget Alerts

Track LLM Spend in n8n—and get budget alerts before costs spike

This n8n workflow automatically monitors your recent executions, extracts LLM token usage, estimates per-call USD cost, logs usage in an n8n Data Table, and sends Slack/email/Telegram/custom webhook alerts when spend goes over budget or suddenly spikes.

What this workflow does

• Runs every 15 minutes on a schedule to keep spend tracking near real-time.
• Fetches execution data from the n8n Public API since the last saved checkpoint.
• Extracts LLM usage from AI-related activity, including token usage from chat model sub-nodes and raw provider responses captured from nodes like OpenAI, Anthropic, Google Gemini, and HTTP Request.
• Calculates token totals and estimates cost in USD using configured per-model pricing, including flags for unpriced models.
• Logs only new LLM-call rows into an n8n Data Table using a unique callKey to deduplicate across runs.
• Analyzes the last 8 days of usage to detect:
  • Budget overruns per workflow
  • Spend spikes vs the prior 7-day daily average
  • Excessive calls per execution (looping behavior)
  • Models missing pricing
• Optionally unpublishes over-budget workflows when configured, then routes alerts to Slack, SMTP email, Telegram, or a custom webhook.

Use cases

• Catch LLM cost overruns in production automations before invoices arrive.
• Detect runaway loops that trigger excessive LLM calls inside n8n executions.
• Monitor provider usage across OpenAI/Anthropic/Gemini—especially when prompts or routing logic changes.

Technical details

• Integrates with n8n Public API (via n8n API key + base URL) to read executions (and optionally unpublish workflows).
• Writes to an n8n Data Table with columns such as callKey, execution/workflow metadata, token counts, costUsd, estimated flags, and finishedAt.
• Uses core n8n nodes including if, switch, set, and code, plus Slack for notifications.